This entry of Machine Readable is by Sam Steyer, Head of Data Science & Head of Customer Success at Halcyon
It’s a safe bet that anyone reading this post works with information. You either create it through your business processes, or absorb it to feed those processes. In fact, you probably both create and consume important information constantly, even if you have not thought of information in a producer-consumer way.
You are simultaneously information’s creator and its captive, and that latter state is a centuries-old pain point in business. It is also something that Halcyon aims to solve: getting from initial engagement with information to durable intelligence about the markets and industries that matter most to you. That means being faster, and clearer, and applying intelligence at a system level that works for everyone.
Early on in scoping our approach, Halcyon’s founding team set out our sense of customer information journeys. It’s something that we - and you - know intuitively but we found it useful to understand it programmatically and sequentially. This understanding of how information travels and changes critically informs what Halcyon is building.
Here is how new information flows:
- Discovery is first. There is usually an inception moment: it could be an event, a conversation, a publication, or a request. It is the kickoff of a process.
- Collection is next. We all have our ways of getting information. Some of these processes are highly rigorous, and some are deliberately informal. All of them serve a purpose: bridging from discovery to input into deeper work.
- Focus follows. This is where a person decides where new information lies within his or her own framework. Does this information serve a short-term need? A long-term strategy? Is it specific to one sector, or does it cut across business priorities? (Focus can also be a matter of priority; we at Halcyon are fond of Eisenhower’s 2x2 matrix of urgency and importance as a way of prioritizing information processing.)
- Analysis follows that. Analysis means transforming information into something relevant for an organization and for specific purposes. It includes validating information, harmonizing it, creating a model that predicts outcomes, consolidating those outcomes, and finally creating an organizational consensus.
- Delivery comes after. Different organizations absorb analysis in different ways (some want a spreadsheet, others bullet points; memos are required in some organizations, and slide decks mandatory in others).
- Outcome is the ultimate goal of information flow. That outcome can be a simple decision with limited scope, or it could be a grand change of strategy.
There are multiple ways to improve the steps above. You can make each of them faster; you can make them higher-volume; you can make them more tailored; you can make them more sharable; you can make them more replicable. Improving any one of those steps confers advantage; improving them all is the ultimate goal.
Artificial intelligence has a distinct role to play in these improvements. Large language models (LLMs) allow us to more easily query unstructured data, and compose text responses. They also allow us to quickly interact with, and iterate on, a thread of ideas. AI capabilities are improving rapidly, and they are also expanding in scope.
Last year, Harvard Business School published an important paper on how AI impacts knowledge worker productivity and work quality. The authors describe current AI as a “jagged technological frontier” in which AI does some tasks very well, but has no capability with tasks of similar difficulty. One of the report authors, Ethan Mollick of Wharton, memorably describes AI as “a somewhat weird alien that wants to work for free for you.” That intern is fast, and it can be nimble too, and at Halcyon we’re putting it to work.
But AI on its own is not everything, in particular for specialized areas of inquiry like rulemaking, regulation, and docketed proceedings. Frontier models do not necessarily train on things like Environmental Protection Agency final rules, or the intricacies of how utility holding companies interact with subsidiaries and joint ventures. And as our Founding Engineer Alexander Huras notes, models are currently very limited in their ability to remember information over time. Like an intern, they need institutional knowledge imparted in lieu of individual memory. They also need guidance, coaching, and a sense that the job they do is worth doing and worth doing well.
Collecting, understanding, and focusing all of those information priorities is what Halcyon does. What we do will allow you to move raw information with greater speed and greater velocity towards informed decisions through an evidence-driven, thoroughly documented process. To echo what we said in our launch post: accelerating the path from information to intelligence becomes an instrument of speed and scale.
Sign up for our newsletter to get all this awesome content in your inbox!